Accession Number:

ADA010931

Title:

Pattern Discrimination Based on an Exponential Error Criterion,

Descriptive Note:

Corporate Author:

RAND CORP SANTA MONICA CA

Personal Author(s):

• Zucker, Joel Steven

Report Date:

1975-02-01

Pagination or Media Count:

109.0

Abstract:

This paper introduces a new error criterion for use in the determination of discriminant functions for pattern recognition. The criterion, which is based on an exponential weighting of discriminant function values, combines the ability to assure the error-free classification of separable training patterns with the desirable extrapolative performance usually associated with classifiers based on the pattern statistics.

Descriptors:

• *PATTERN RECOGNITION
• WAVEFORMS
• PROBABILITY DENSITY FUNCTIONS
• SPEECH RECOGNITION
• BAYES THEOREM
• DISCRIMINATE ANALYSIS
• ELECTROCARDIOGRAPHY
• THEOREMS

Subject Categories:

• Cybernetics
• Statistics and Probability

Distribution Statement:

APPROVED FOR PUBLIC RELEASE